Description

Question 1

Train a multi-layer perceptron using backpropagation for the following two dimensional dataset. We have provided some code to implement the MLP, but anything missing you will need to code up yourself (independently – without using online code sources) and add it to the code. There are two python files, one with the MLP implementation (labeled mlp.py) and another for instantiating the MLP and using it (run.py). If all the proper code is added in the correct manner, only run.py needs to be run. MLP.py should NOT be executed directly. It is a module you import into your run.py file.

Figure 1: 2-D dataset

For the report, use confusion matrices to demonstrate the accuracy of your MLP, and discuss the methods you used to achieve the highest accuracy. How many nodes do you have on your hidden layer? How did changing the learning rate change your results? What role does increased different number of iterations have on accuracy?